A review of terrain detection systems for applications in locomotion assistance

The article describes the process of development of an essentially new wheel suitable both for moving on flat ground and for travelling on stairs. The stair-climbing wheel is composed of rotary cir...

https://journals.sagepub.com/doi/pdf/10.1177/1729881417749470

The synthesis of a segmented stair-climbing wheel:

Abstract Staircases are some of the most common building structures in urban environments. Stair detection is an important task for various applications, including the environmental perception of exoskeleton robots, humanoid robots, and rescue robots and the navigation of visually impaired people. Most existing stair detection algorithms have difficulty dealing with the diversity of stair structure materials, extreme light and serious occlusion. Inspired by human perception, we propose an end-to-end method based on deep learning. Specifically, we treat the process of stair line detection as a multitask involving coarse-grained semantic segmentation and object detection. The input images are divided into cells, and a simple neural network is used to judge whether each cell contains stair lines. For cells containing stair lines, the locations of the stair lines relative to each cell are regressed. Extensive experiments on our dataset show that our method can achieve 81.49 $$\%$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mo>%</mml:mo> </mml:math> accuracy, 81.91 $$\%$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mo>%</mml:mo> </mml:math> recall and 12.48 ms runtime, and our method has higher performance in terms of both speed and accuracy than previous methods. A lightweight version can even achieve 300+ frames per second with the same resolution. 

https://www.nature.com/articles/s41598-022-20667-w.pdf

Deep leaning-based ultra-fast stair detection

The evolutionary process of modern robotics helps the military forces to revamp their aged systems with today's advanced technology. For defending any kind of criminal activities like hostage situation created by any extremist or a militant group the uses of all terrain military assistance robot has increased from past time. In this paper we have approached a stair climbing robot with some creative features which can assist a military group to complete a mission or achieve any milestone. We have focused on reliable designs and unique stair climbing mechanisms with the detection of stairs through image processing. Moreover detecting the human movement and different kind of guns using the image processing algorithms helps us to upgrade our system into a semi-autonomous robot. Using radio frequency communication protocol, wide range of working space is achieved which helps the user to control the robot from a safe distance from the crime zone. Experimental analysis of the system justify the reliability as well as the efficiency of the design and the algorithms for practical scenario.

A Semi-Autonomous Tracked Robot Detection of Gun and Human Movement Using Haar Cascade Classifier for Military Application

In medical contexts, the use of assistive exoskeletons for the rehabilitation of people with impaired mobility represents a common practice. Recent advances suggest that, soon, such mechatronic systems will also be used to assist people in their everyday life. In order to reach such target, exoskeletons must become able to perceive the environment. To this purpose, a system for the parametric identification of a staircase is proposed in this article. More precisely, given a staircase of unknown geometry, the system identifies its 3D shape. Furthermore, it also estimates the reciprocal orientation and distance between the exoskeleton and the staircase. Differently from other approaches, this result is achieved by means of low cost devices: an inertial measurement unit, two ranging sensors, and an Arm-Cortex processor. Starting from the ranging sensors acquisitions, the staircase model is identified in real time, during the execution of a step. The proposed procedure is based on an extended recursive total least squares strategy, in order to fully exploit the computational capabilities of the Arm processor, and it is characterized by execution times smaller than $10^{-3}$ s. The estimation algorithm has been tested on an actual exoskeleton and the resulting experimental outcomes are compared with the results obtained through alternative methods.

/pdf/a-real-time-3d-reconstruction-of-staircases-for-fh8skcfv51.pdf

A Real-Time 3D Reconstruction of Staircases for Rehabilitative Exoskeletons

This paper presents a novel stair recognition method by utilizing three-dimensional (3D) scanning for a hexapod robot to climb stairs. The method is based on the premise that stairs are composed of several parallel planar surfaces. A 3D laser range finder (LRF) system equipped on a hexapod robot is developed for performing 3D scanning. The algorithm can be divided into two procedures. At first, detecting the edge points of stair tread surfaces is performed in 2D range images. In order to enhance the robustness of the method and extract the edges more accurately, edge point searching is executed between neighbouring scan planes. Next, the extraction of the tread surfaces is carried out by using these detected edge points in 3D space. The performance of the proposed method is investigated through various experiments. As the experimental results demonstrate, the method is sufficient accurate to detect the tread surfaces and useful for a hexapod robot to climb the stair.

Zhiheng Zheng

Papers

A method to detect stairs with three-dimensional scanning for hexapod robot stair climbing